Anthropic has abandoned its most important principle: it will no longer pause dangerous models if competitors release them first

Anthropic has significantly altered its foundational "Responsible Scaling Policy," a move with considerable implications for the AI industry and its commitment to safety. Previously, the company pledged to halt or delay the development of potentially dangerous AI models if safeguards could not keep pace with their advancements. This policy change signals a shift away from automatic braking mechanisms towards a framework of flexible public commitments and third-party risk audits. The core of this revision involves Anthropic's decision to no longer unilaterally pause model development when competitors advance without similar restrictions. This strategic adjustment acknowledges the intense competitive landscape, with major players like OpenAI, Google, and xAI pushing forward rapidly. Furthermore, the company highlights political pressures and the US approach to AI regulation as factors influencing this policy evolution, suggesting a move toward a less restrictive development environment. This recalibration raises questions about the future of AI safety standards, particularly in contrast to Europe's more regulatory approach. While Anthropic now emphasizes public commitments and external audits, the removal of its internal, automatic safety brake could accelerate AI development at the potential expense of robust, pre-emptive risk mitigation. The long-term impact on public trust and the global AI governance debate remains to be seen as other entities navigate this evolving landscape.